Method and apparatus of recognizing facial expression base on multi-modal

摘要

Constructing a CNN model expression database through learning using the CNN model; Constructing an SVM model expression database through learning using the SVM model; Obtaining facial expression data using the CNN model for the recognition image; Acquiring facial expression data using an SVM model for a recognition image; The CNN model facial expression database is searched for the facial expression data using the CNN model to extract the closest CNN model facial expression classification data, and the first weight is added to the extracted CNN model facial expression classification data to obtain CNN model weight operation expression data Calculating; The SVM model facial expression database is searched for the facial expression data using the SVM model to extract the closest SVM model facial expression classification data value and the second weight is added to the extracted SVM model facial expression classification data to obtain the SVM model weight operation facial expression data value ; A facial recognition step of combining the CNN model weighting operation facial expression data and the SVM model weighting facial expression data to calculate final facial expression data and selecting a most probable value to recognize facial expression; The method comprising the steps of: